Aetina CoreEdge MXM Series: Industrial AI Computing Enters the Blackwell Era

Aetina Corporation has unveiled its CoreEdge MXM AI Accelerator Module series, leveraging NVIDIA’s RTX PRO Blackwell Embedded GPUs to deliver up to 3X AI performance improvements over previous generations. The industrial-grade modules are engineered specifically for deployment in autonomous mobile robots (AMRs), unmanned vehicles, smart manufacturing systems, and medical equipment where reliability and computational density are paramount.

Architecture and Performance Metrics

The flagship MX5000B-XA module integrates the NVIDIA RTX PRO 5000 Blackwell GPU, featuring 10,496 CUDA cores, 80 RT cores, and 320 Tensor cores. The module achieves 40.62 TFLOPS in FP32 operations and demonstrates significant performance gains in practical workloads: validation testing shows 2X faster inference on the Gemma-2-9b large language model and an 80% improvement in Geekbench AI benchmarks compared to predecessor architectures.

Key technological advances include support for FP4 precision compute, DLSS 4 multi-frame generation, and GDDR7 memory delivering bandwidth up to 896 GB/s. These capabilities enable deployment of computationally intensive models including large language models and vision transformers in power- and space-constrained edge environments, a critical requirement for physical AI applications demanding low-latency, high-throughput processing.

Industrial Design Philosophy

Adhering to the MXM 3.1 industrial standard, the CoreEdge modules occupy approximately one-fifth the board space of conventional graphics cards while maintaining workstation-class computational performance. The modules incorporate several industrial-hardening features distinguishing them from commercial products:

• Extended temperature range: -40°C to +85°C operational envelope
• Environmental protection: Conformal coating for dust and moisture resistance
• Mechanical robustness: High-vibration resistant mounting holes on MXM Type B variants
• Power flexibility: Adjustable TGP settings for diverse thermal envelopes
• Lifecycle support: Five-year product availability guarantee

Aetina’s engineering validation process includes wide-temperature component selection, extensive thermal cycling, cold boot verification, and 100% environmental chamber testing prior to shipment. Evaluation kits featuring PCIe Gen4 x16 interfaces and HDMI outputs are available to expedite integration workflows.

Market Positioning and Availability

According to Jackal Chen, Director of AI Accelerator Card & GPU Products Division at Aetina, the series represents preparation for multimodal AI agents and real-time accelerated computing demands. The company offers value-added engineering services to reduce system integration complexity and time-to-market.

The MX5000B-XA and MX500B-QA modules are currently shipping. Mid-range MX4000B and MX2000B variants are scheduled for mass production in Q1 2026, providing scalable options across performance tiers.

Technical Implications

The CoreEdge series addresses a critical inflection point in edge AI deployment: the need to execute increasingly sophisticated models in resource-constrained, reliability-critical environments. By delivering data center-class AI capabilities in the MXM form factor with industrial environmental ratings, these modules enable new architectures for autonomous systems, manufacturing automation, and medical devices where computational power cannot compromise safety, reliability, or physical constraints.

For detailed specifications and integration documentation, visit http://www.aetina.com